North Slope Borough, Alaska · 70.25°N · 148.34°W · Market: Telecom

Telecom towers and wind-induced vibrations

Tower natural frequency
~1.5 Hz1
Lock-in wind window
13-21 mph1
Peak recorded wind
109 mph2
Tower failures since 1959
1403

What is wind-induced vibration, and where does the damage land?

Steady wind sheds vortices off alternate sides of a slender tower, pushing it sideways. The tower sways across the wind, rocks along it, and can twist. When the shedding rate meets a natural frequency, the motion locks in and grows. Bending stress is highest at the base, so that is where the fatigue damage concentrates.

wind (E)Antenna mount6% of service lifeUpper mast section12% of service lifeMid-mast splice23% of service lifeLower flange splice44% of service lifeBase plate weld67% of service lifehow it moves (top view)side to side (cross-wind)back & forth (along-wind) · twist
under 25% life25-60%over 60%: focus areaCross-wind sway dominates; stress and fatigue peak at the base. Schematic, illustrative.

At what wind speed does my structure lock in?

Drag the wind speed to see predicted amplitude. The response is a band, because real motion scatters around the trend. This is the prediction; the sensors below record what happens.

1st mode lock-in2nd mode lock-in010203040vibration amplitude (mm)010203040506070perpendicular wind speed (mph)
Perpendicular wind
17 mph
Lock-in: resonance risk
Shedding has synchronized with a structural mode. Amplitude swells into the resonance hump and fatigue cycles pile up fast.
Predicted amplitude
9.1-17.7 mm
Shedding frequency
1.52 Hz
1st / 2nd mode
1.5 / 4.5 Hz
Lock-in winds
17, 50 mph

How much fatigue life has each location already spent?

Measured cycles are sorted by amplitude and summed against an S-N curve to give each location's spent fatigue life. The ranking drives the real decision: which sections need a damper, and how many to buy.

Fatigue spent, by location
Base plate weld
147M cycles · peak 32 mm · 67% of service life
1010¹10²10³1010101010cycles counted (log scale)0.512481216242832amplitude bucket (mm)

Cycles are binned by amplitude, converted to a stress range, and summed against an S-N curve by Miner's rule. The red high-amplitude buckets, though rare, carry most of the damage. Illustrative synthetic data, not a measurement of any real structure.

How long the steel lasts depends on material and detail
2050100200500stress range (MPa)10³101010101010cycles to failureendurance limit (steel)PlainWeldedAluminium
Plain steelWelded steel (Cat 90)AluminiumSteel flattens to an endurance limit; aluminium keeps falling. Illustrative.

Damage is summed against the EN 1993-1-9 Category 90 welded-detail curve.4

Where do you put the sensors?

Three sensors, each answering a different question. Near the top for the cleanest motion signal, mid-mast to catch the higher mode, and at the base where stress and fatigue are highest. Placement follows the tower's mode shapes.

Communications towers on a snow-covered mountain summit
Deborah Lee Soltesz, public domain (CC0 1.0), via Wikimedia Commons.
Where the sensors go, top to base
Near the antenna mount
Maximum sway: the clearest signal for detecting motion and identifying mode 1
Mid-mast
Higher-mode antinode: separates the 2nd bending mode and torsion
Base / foundation weld
Maximum bending stress: anchors the fatigue estimate where cracks start

Which sensors handle vibration monitoring?

Wind-induced vibration is a fast, three-axis signal at a remote, frozen, off-grid site. The sensor has to capture all of it and phone home on its own power.

Omnidots SWARM triaxial vibration monitor
Omnidots SWARM, a triaxial vibration monitor. See it in the Market.
What a WIV sensor needs
Triaxial
Separates cross-wind sway, along-wind rocking, and torsion.
1 kHz capture
Fast enough to resolve the waveform, so vortex shedding is told apart from turbulence.
Cold and ice rated
Runs through the Arctic winter and survives icing on the housing.
Wireless backhaul
Cellular telemetry for sites with no fixed network.
Solar plus battery
Self-powered for sites with no grid, through the polar night.

What wind drives the vibration, and from which direction?

The damaging events arrive from a narrow bearing: the prevailing E wind, which hits the tower most broadside. The rose shows how often it blows; the scatter shows how hard it shakes.

4%8%12%16%NESW

Wind blows from the E most of the year at Deadhorse, mean 11.9 mph, the direction most nearly perpendicular to the tower face. That is why the base weld is the focus area.

Wind speed (mph)
31+
19-31
12-19
7-12
0-7

Petal length is the share of hours from each direction; segments split it by speed band. Shape anchored to Deadhorse PASC climatology; per-direction values illustrative.

Amplitude by wind direction
0481115amplitude (mm)090180270360wind direction (°, 0 = N)

What happens if we install a tuned mass damper?

A wind forecast run through the model shows expected amplitude with and without a damper. Because fatigue scales with stress cubed, a modest cut in amplitude sharply slows the fatigue clock.

0481115predicted amplitude (mm)Day 1Day 2Day 3Day 4Day 5Day 6Day 7
as-is (bare) with tuned damper
Forecast peak, as-is
13 mm
Forecast peak, with damper
5 mm
Fatigue accrual cut
10×

The damper raises modal damping from 0.8% to 2.2% of critical, flattening the resonant peak and trimming the broadband response every day. Because fatigue scales with stress cubed, the accrual rate drops far more than the amplitude. Illustrative forecast.

Is the vibration vortex shedding, or turbulence?

The distinction sets the fix. Vortex shedding is a narrow oscillation across the wind at the natural frequency, so the cross-wind axis dominates one sharp spectral peak. Turbulence spreads energy across every axis and frequency.

Xcross-windYin-lineZaxial
Verdict: vortex-induced lock-in
Dominant frequency 1.5 Hz, the structure's 1st mode, so this is lock-in. The cross-wind axis carries 2.4× the energy of the in-line axis: a narrow, across-flow oscillation is the signature of vortex shedding. Random gust buffeting would spread energy evenly across both axes and the spectrum.
1,000 Hz · 3-axis · 4s window (4,000 samples/axis)Illustrative waveform.
The same verdict, read from the amplitude distribution
0234669samples-1.1401-0.7551-0.37010.01490.39990.7849cross-wind acceleration (sample trace)

summarize_dataset on the cross-wind trace above, plotted straight from its histogram JSON. The U-shape — samples piling up at the two extremes — is the signature of a dominant sinusoid sweeping through its peaks: lock-in. Pure turbulence would mound up in the middle instead. Synthetic sample (see Sources).

Are the sensors healthy?

Remote Arctic monitoring is only as good as its uptime. Battery state, last-seen time, and link status sit on one panel, so a unit going dark in a freeze-up is caught before it leaves a gap in the fatigue record, not discovered at the next site visit.

Sensor fleet3/5 online
BX-01Antenna mount
Streaming 1 kHz, 3-axis · seen 2 min ago
batt91%
Online
BX-04Upper mast section
Streaming 1 kHz, 3-axis · seen 3 min ago
batt78%
Online
BX-07Mid-mast splice
Battery low: solar input down since freeze-up · seen 5 min ago
batt14%
Service soon
BX-09Lower flange splice
Streaming 1 kHz, 3-axis · seen 4 min ago
batt64%
Online
BX-12Base plate weld
No telemetry: likely comms or power fault · seen 9 days ago
batt0%
Offline

What are conditions at the site right now?

Monitoring an Arctic site means staring down the same weather the structure does: wind, light, and storms, in real time.

Live camera: Dalton Highway, Atigun Pass, Brooks Range

Live: Dalton Highway, Atigun Pass, Brooks Range (Alaska DOT&PF 511).

Live wind over Prudhoe Bay (Windy).

Why is measuring vibration harder in the Arctic?

The wind never lets up

Deadhorse averages 11.9 mph from the E and has gusted to 109 mph. Cyclic loading is near-constant, so you have to capture events continuously, not sample them.5

No grid, thin comms

North Slope sites have no line power and little bandwidth. SSD instrumented more than 30,000 aboveground spans in exactly these conditions; historically the data came off by hand on infrequent visits.6

Cold and ice fight the instruments

Freeze-up cuts solar input and saps batteries, and ice loads both sensors and structure. Ice is implicated in most of the 140 US tower failures CRREL has recorded since 1959.3

No standard told anyone to look

TIA-222-H carried no method to assess vortex-induced vibration. Only the 2024 TIA-222-I adds vibration and fatigue, so legacy towers were never analyzed for it.7

So how do you know the tower is still safe?

About 140 US communication towers have collapsed under ice and wind since 1959, roughly four a year, and the count is acknowledged to be incomplete.3 Each one is filed under the weather on the day it falls, not the years of cyclic loading that spent its fatigue life first. A calendar inspection cannot tell you the natural frequency has drifted or a connection has loosened. Continuous monitoring can, and it can warn you before the next big event.

Live alerts

Lock-in forecast, ~14 h: 18-22 mph from the E drives the base weld to a predicted 16 mm (+0.4% fatigue). The alarm fires before the wind arrives; an inspection would miss it.

Act now

Can I point my own AI agent at this over MCP?

Yes. The same general BeadedCloud MCP tools run on vibration and fatigue data — no key needed. Below is live output from the fatigue tool on this page's amplitude histograms; an agent calls the identical tool on your measured traces.

Ask BeadedCloud over MCPpreview

Tap a question below to see what a connected agent returns ↓

Preview on this page's synthetic sample (see Sources) — the tools are real and run on your measured data.Set up MCP →
Real bearing run-to-failure: the trend an agent gets back
00.20.40.60.811.21.41.6vibration RMS (accel.)Jun 20 17:00Jun 21 16:00Jun 22 16:00Jun 23 15:00Jun 24 14:00Jun 25 13:00ch1 accelch2 accel
ch1 accelch2 accelPlotted straight from analyze_vibration's trend JSON — real run-to-failure data, condition CRITICAL by the final snapshot (Mendeley 5hcdd3tdvb, CC-BY-4.0).
Spec sheet

Take the engineering spec sheet with you

One page: what the monitoring system measures, the sensor and deployment specs, and the standards it aligns to. PDF.

We will add you to the early-access list. No spam.

Want this view on your structure?

Circumpolar.ai turns vibration and wind sensors into live amplitude, frequency-drift, and fatigue-cycle dashboards with alarms, built for sites with no grid power and thin connectivity. Early access is open.

Sign up for circumpolar.ai